Implement generative AI and agentic solutions 2.2 Build agents by using Foundry

You're moving an orchestrator agent from the classic Agents API to the current Foundry…

Study guide skill: Implement orchestrated multi-agent solutions

Question B2-015

You're moving an orchestrator agent from the classic Agents API to the current Foundry Agent Service. It must delegate parts of each request to a partner's agent that runs outside Foundry and exposes an A2A-compatible endpoint, while your agent keeps managing the conversation. What should you use?

Select one answer.

  • A The file search tool with the partner's documentation
  • B The Connected Agents tool
  • C Function calling that names the partner's model
  • D The A2A tool (a2a tool type) with a project connection to the partner's endpoint ✓ Correct answer

Answer: D

The A2A tool connects a Foundry agent to any A2A-compatible endpoint through a standard protocol: the remote agent processes the request, and your agent uses the response and keeps managing the conversation. The generally available a2a tool type is recommended for new integrations. The Connected Agents tool from the classic API isn't available in the new Foundry Agent Service, and file search or function calling doesn't delegate to another agent.

Deep dive

How it works

The A2A tool connects a Foundry agent to a remote endpoint that speaks the Agent2Agent (A2A) protocol, so your agent can exchange context and collaborate with external agents through a standard protocol. The division of labor is exactly what the question asks for: the remote agent processes the request and returns a response, and your Foundry agent uses that response to answer the user and keeps managing the conversation.

Setup in brief:

  1. Create a project connection for the partner's A2A endpoint. It stores authentication securely and lets you reuse it across agent versions.
  2. Add the tool with the a2a tool type. It's the generally available option and recommended for new integrations; the older a2a_preview type stays in preview for existing ones.
  3. Describe in the instructions when to delegate.

Migration context: the Connected Agents tool from the classic Agents API isn't available in the new Foundry Agent Service. The replacements are the A2A tool, for any A2A-compatible endpoint, or workflows, for orchestrating several Foundry agents declaratively.

The other options, and when you'd use them

  • File search with the partner's documentation: File search retrieves information from your documents. It reads content; it doesn't delegate work to another agent.
  • The Connected Agents tool: It isn't available in the new Foundry Agent Service. The migration table's recommendation for it is the A2A tool.
  • Function calling that names the partner's model: With function calling, the model asks your app to call a function, and your app executes it. That's your own code, not an A2A delegation to a partner's agent.

Exam tip

Delegate to an external agent with an A2A endpoint = A2A tool (a2a type) + project connection. Connected Agents is classic-only.

References

Last reviewed:

Ready to test yourself? The quiz asks this and every other AI-103 question, with instant feedback. Practice this in a quiz →

All Generative AI & agents questions · AI-103 study guide

Thanks for your feedback!